effective-html-plan
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to ingest untrusted source material (conversations, project instructions, plans) and transform it into self-contained HTML files. This could allow malicious content within the source material to influence agent behavior or inject scripts into the rendered output. 1. Ingestion points: SKILL.md directs the agent to read conversation history, project instructions, and user-supplied plans. 2. Boundary markers: The instructions do not define explicit delimiters or headers to isolate external data from instructions. 3. Capability inventory: The skill leverages the agent's capability to generate and write HTML files to the local filesystem and return absolute paths. 4. Sanitization: No instructions are provided for escaping or sanitizing the source content before it is embedded in the HTML.
- [EXTERNAL_DOWNLOADS]: MIRROR.md references an upstream repository at github.com/plannotator/effective-html.git as the source for the skill's content and versioning.
- [NO_CODE]: The skill consists solely of markdown instructions, a license, and configuration files, containing no executable scripts, binaries, or logic in standard programming languages.
Audit Metadata